
import time
start_time = time.time()


import numpy as np
import pandas as pd
import os, time, pickle, gzip

import matplotlib.pyplot as plt
import seaborn as sns
color = sns.color_palette()
import matplotlib as mpl

from sklearn import preprocessing as pp

%matplotlib inline


TRX = pd.read_csv('C:\WORK/mnistTRX.csv')
VLX = pd.read_csv('C:\WORK/mnistVLX.csv')
TSX = pd.read_csv('C:\WORK/mnistTSX.csv')

TRy = pd.read_csv('C:\WORK/mnistTRy.csv')['y']
VLy = pd.read_csv('C:\WORK/mnistVLy.csv')['y']
TSy = pd.read_csv('C:\WORK/mnistTSy.csv')['y']

iTR = range(0, len(TRX))
iVL = range(len(TRX), len(TRX)+len(VLX))
iTS = range(len(TRX)+len(VLX), len(TRX)+len(VLX)+len(TSX))

nX = [TRX.shape, VLX.shape, TSX.shape]
ny = [TRy.shape, VLy.shape, TSy.shape]
pd.DataFrame(nX, index=['TRX', 'VLX', 'TSX'], columns=['n', 'p'])


pd.DataFrame(ny, index=['TRy', 'VLy', 'TSy'], columns=['n'])


TRX.describe()


TRy.head()

0    5
1    0
2    4
3    1
4    9
Name: y, dtype: int64


TRy.head()

0    5
1    0
2    4
3    1
4    9
Name: y, dtype: int64


def view_digit(example):
    label = TRy.loc[example]
    image = TRX.loc[example, :].values.reshape([28, 28])
    plt.title('Example: %d Label: %d' % (example,label))
    plt.imshow(image, cmap=plt.get_cmap('gray'))
    plt.show()


view_digit(0)


from sklearn.preprocessing import OneHotEncoder, LabelBinarizer
LB = LabelBinarizer()
TRY = LB.fit_transform(TRy)
TRY[0]

array([0, 0, 0, 0, 0, 1, 0, 0, 0, 0])


TRXO = TRX.groupby(TRy).apply(lambda x: x.sample(100, random_state=2018))
iTRO = TRXO.index.to_frame()[1]
TRXO.set_index(iTRO, inplace=True)
TRy0 = TRy.iloc[iTRO]


from sklearn.decomposition import PCA

n_components = 784
whiten = False
random_state = 2018

Epca = PCA(n_components=n_components,
          whiten = whiten,
          random_state=random_state)
TRXTpca = Epca.fit_transform(TRX)
TRXTpca = pd.DataFrame(data=TRXTpca, index=iTR)


TRXTpca.corr().round(5)


TRXTpca.var().round(5)

0      5.10829
1      3.70098
2      3.25868
3      2.82008
4      2.54673
        ...   
779    0.00000
780    0.00000
781    0.00000
782    0.00000
783    0.00000
Length: 784, dtype: float64


TRXTpca.cov().round(5)


def summaryPCA(PCAobj):
    eigval = pd.DataFrame(PCAobj.explained_variance_)
    prop = pd.DataFrame(PCAobj.explained_variance_ratio_)
    cumul = pd.DataFrame(np.cumsum(prop)/prop.sum())
    vrexp = pd.concat([eigval, prop, cumul], axis=1).T
    vrexp.index = ['Eigenvalue', 'Prop', 'Cumulative']
    return vrexp
vrexp = summaryPCA(Epca)
vrexp.loc[:, [0, 1, 2, 9, 19, 49, 99, 199, 299, 399, 499, 599, 699]].round(4)


fig, ax = plt.subplots(ncols=3, figsize=(16, 4))

ev = sns.barplot(data=pd.DataFrame(vrexp.loc['Eigenvalue', :20]).T, color='b', ax=ax[0]);
ev.set(title='Eigenvalues of S', xlabel='PC', ylabel='Eigenvalues')
pr = sns.barplot(data=pd.DataFrame(vrexp.loc['Prop', :20]).T, color='b', ax = ax[1]);
pr.set(title='Variance Explained by PC', xlabel='PC', ylabel='Proportion')
cm = sns.barplot(data=pd.DataFrame(vrexp.loc['Cumulative', :20]).T, color='b', ax=ax[2]);
cm.set(title='Variance Explained by PC(Cumulative)', xlabel='PC', ylabel='Cumulative Proportion', ylim=[0, 1]);


def scatterPlot(xDF, yDF, algoName):
    tempDF = pd.DataFrame(data=xDF.loc[:, 0:1], index=xDF.index)
    tempDF = pd.concat((tempDF, yDF), axis=1, join="inner")
    tempDF.columns = ["First Vector", "Second Vector", "Label"]
    sns.lmplot(x="First Vector", y="Second Vector", hue="Label", data=tempDF, fit_reg=False)
    ax = plt.gca()
    ax.set_title("Separation of Observations using"+algoName)


scatterPlot(TRXTpca, TRy, 'PCA')


fig, axs = plt.subplots(nrows=2, ncols=2, figsize=(12,12))
d0 = sns.scatterplot(x=TRXTpca.loc[:, 0], y=TRXTpca.loc[:, 1], hue=TRy==0, alpha=0.25, ax=axs[0,0]);
d0.set(xlabel='PC1', ylabel='PC2', title='0')
d1 = sns.scatterplot(x=TRXTpca.loc[:, 0], y=TRXTpca.loc[:, 1], hue=TRy==1, alpha=0.25, ax=axs[0,1]);
d1.set(xlabel='PC1', ylabel='PC2', title='1')
d5 = sns.scatterplot(x=TRXTpca.loc[:, 0], y=TRXTpca.loc[:, 1], hue=TRy==5, alpha=0.25, ax=axs[1,0]);
d5.set(xlabel='PC1', ylabel='PC2', title='5')
dd = sns.scatterplot(x=TRXTpca.loc[:, 0], y=TRXTpca.loc[:, 1], hue=TRy, alpha=0.25, ax=axs[1,1]);
dd.set(xlabel='PC1', ylabel='PC2', title='0~9')

[Text(0.5, 0, 'PC1'), Text(0, 0.5, 'PC2'), Text(0.5, 1.0, '0~9')]


TRX.iloc[:, [350,406]].head()


XX = pd.DataFrame(data=TRX.iloc[:, [350,406]], index=iTR)
XX = pd.concat([XX, TRy], axis=1, join='inner')
sns.lmplot(x='x351', y='x407', hue='y', data=XX, fit_reg=False)
ax = plt.gca()
ax.set_title('Separation of Observations Using Original Feature Set (x351, x407);')

Text(0.5, 1.0, 'Separation of Observations Using Original Feature Set (x351, x407);')


from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure(figsize=(16,9))
ax = plt.axes(projection='3d')
p = ax.scatter3D(TRXTpca[0], TRXTpca[1], TRXTpca[2], alpha=0.25, c=TRy)
ax.set_xlabel('X:PC1')
ax.set_ylabel('Y:PC2')
ax.set_zlabel('Z:PC3')

Text(0.5, 0, 'Z:PC3')


import plotly.express as px


fig = px.scatter_3d(TRXTpca, x=0, y=1, z=2, color=TRy, size_max=10)
fig.show()


TRXTO = TRXTpca.groupby(TRy).apply(lambda x :x.sample(100, random_state=2018))
iTRO = TRXTO.index.to_frame()[1]
TRXTO.set_index(iTRO, inplace=True)
TRy0 = TRy.iloc[iTRO]

fig = px.scatter_3d(TRXTO, x=0, y=1, z=2,
                   color = TRy0==1,
                   symbol=TRy0==1,
                   size=1+(TRy0==1).astype(int),
                   size_max=9,
                   opacity=0.5)
fig.show()


from sklearn.model_selection import KFold, StratifiedKFold, GridSearchCV
from sklearn.metrics import mean_squared_error

KF5 = KFold(n_splits=5, shuffle=True, random_state=2018)

def negmse(estimator, X, y=None):
    XT = estimator.transform(X)
    Xh = estimator.inverse_transform(XT)
    return -1* mean_squared_error(X, Xh)


parampca = {'n_components': [100, 200, 300, 400, 500]}

Epca2 = PCA()
GSpca = GridSearchCV(Epca2, param_grid=parampca, cv=KF5, scoring=negmse)
%time GSpca.fit(TRX)
GSpca.score(TRX)

Wall time: 3min 37s

-5.0286759245809235e-05


cvresult = pd.DataFrame(GSpca.cv_results_)
cvresult


sns.barplot(x='param_n_components', y='mean_test_score', data=cvresult, color='b')

<AxesSubplot:xlabel='param_n_components', ylabel='mean_test_score'>


print(GSpca.best_score_.round(4))
print(GSpca.best_params_)
print(GSpca.best_estimator_)

-0.0001
{'n_components': 500}
PCA(n_components=500)


from sklearn.decomposition import IncrementalPCA

n_components = 784
batch_size = None

Eipca = IncrementalPCA(n_components=n_components,
                      batch_size=batch_size)

%time TRXTipca = Eipca.fit_transform(TRX)
TRXTipca = pd.DataFrame(data=TRXTipca, index=iTR)

VLXTipca = Eipca.transform(VLX)
VLXTIPCA = pd.DataFrame(data=VLXTipca, index=iVL)

print(TRXTipca.shape)
scatterPlot(TRXTipca, TRy, 'lncremental PCA')

Wall time: 9.83 s
(50000, 784)


from sklearn.decomposition import SparsePCA

n_components = 100
alpha = 0.0001
random_state = 2018
n_jobs = -1

Espca = SparsePCA(n_components = n_components,
                 alpha = alpha,
                 random_state = random_state,
                 n_jobs = n_jobs)
%time Espca.fit(TRX.loc[:10000, :])
%time TRXTspca = Espca.transform(TRX)
TRXTspca = pd.DataFrame(data=TRXTspca, index=iTR)

VLXTspca = Espca.transform(VLX)
VLXTspca = pd.DataFrame(data=VLXTspca, index=iVL)

print(TRXTspca.shape)
scatterPlot(TRXTspca, TRy, 'Sparse PCA')

Wall time: 31.1 s
Wall time: 499 ms
(50000, 100)


from sklearn.decomposition import KernelPCA

n_components = 100
kernel = 'rbf'
gamma = None
random_state = 2018
n_jobs = 1

Ekpca = KernelPCA(n_components=n_components,
                 kernel=kernel,
                 gamma=gamma,
                 n_jobs=n_jobs,
                 random_state=random_state)
%time Ekpca.fit(TRXO)
%time TRXTkpca = Ekpca.transform(TRX)
TRXTkpca = pd.DataFrame(data=TRXTkpca, index=iTR)

VLXTkpca = Ekpca.transform(VLX)
VLXTkpca = pd.DataFrame(data=VLXTkpca, index=iVL)

print(TRXTkpca.shape)
scatterPlot(TRXTkpca, TRy, 'Kernel PCA')

Wall time: 218 ms
Wall time: 2.87 s
(50000, 100)


paramkpca = {
    'gamma':np.linspace(0.01, 0.05, 5),
    'kernel': ['rbf', 'linear']
}
Ekpca2 = KernelPCA(n_components=100, fit_inverse_transform=True, n_jobs=n_jobs, random_state=random_state)
GSkpca = GridSearchCV(Ekpca2, param_grid=paramkpca, cv=3, scoring=negmse)
%time GSkpca.fit(TRXO)
GSkpca.score(TRXO)

Wall time: 4.68 s

-0.05083305496936513


cvresult = pd.DataFrame(GSkpca.cv_results_)
cvresult


sns.barplot(x='param_gamma', y='mean_test_score', hue='param_kernel', data=cvresult);


print(GSkpca.best_score_.round(4))
print(GSkpca.best_params_)
print(GSkpca.best_estimator_)

-0.055
{'gamma': 0.01, 'kernel': 'linear'}
KernelPCA(fit_inverse_transform=True, gamma=0.01, n_components=100, n_jobs=1,
          random_state=2018)


from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline
clf = Pipeline([('kpca', KernelPCA()),
               ('glm', LogisticRegression())])

SKF5 = StratifiedKFold(n_splits=5, shuffle=True, random_state=2018)
param_grid = {
    'kpca__n_components':[100, 200],
    'kpca__gamma': np.linspace(0.01, 0.05, 5),
    'kpca__kernel': ['rbf', 'linear', 'sigmoid']
}

grid_search = GridSearchCV(clf, param_grid, cv=SKF5)
%time grid_search.fit(TRXO, TRy0)
print(grid_search.best_score_)
print(grid_search.best_params_)

Wall time: 33.8 s
0.873
{'kpca__gamma': 0.03, 'kpca__kernel': 'rbf', 'kpca__n_components': 200}


from sklearn.decomposition import TruncatedSVD

n_components = 200
algorithm = 'randomized'
n_iter = 5
random_state = 2018

Etsvd = TruncatedSVD(n_components = n_components,
                    algorithm = algorithm,
                    n_iter = n_iter, 
                    random_state = random_state)

%time TRXTtsvd = Etsvd.fit_transform(TRX)
TRXTtsvd = pd.DataFrame(data=TRXTtsvd, index=iTR)

VLXTtsvd = Etsvd.transform(VLX)
VLXTtsvd = pd.DataFrame(data=VLXTtsvd, index=iVL)

print(TRXTtsvd.shape)
scatterPlot(TRXTtsvd, TRy, 'Truncated SVD')

Wall time: 6.92 s
(50000, 200)


from sklearn.random_projection import GaussianRandomProjection
n_components = 'auto'
eps = 0.5
random_state = 2018

Egrp = GaussianRandomProjection(n_components=n_components,
                               eps=eps,
                               random_state=random_state)

%time TRXTgrp = Egrp.fit_transform(TRX)
TRXTgrp = pd.DataFrame(data=TRXTgrp, index=iTR)

VLXTgrp = Egrp.transform(VLX)
VLXTgrp = pd.DataFrame(data=VLXTgrp, index=iVL)
                       
print(TRXTgrp.shape)
scatterPlot(TRXTgrp, TRy, 'Gaussian Random Projection')

Wall time: 670 ms
(50000, 519)


from sklearn.random_projection import SparseRandomProjection

n_components = 'auto'
density = 'auto'
eps = 0.5
dense_output = False
random_state = 2018
Esrp = SparseRandomProjection(n_components = n_components,
                             density = density,
                             eps = eps, 
                             dense_output = dense_output,
                             random_state=random_state)

%time TRXTsrp = Esrp.fit_transform(TRX)
TRXTsrp = pd.DataFrame(data=TRXTsrp, index=iTR)

VLXTsrp = Esrp.transform(VLX)
VLXTsrp = pd.DataFrame(data=VLXTsrp, index=iVL)

print(TRXTsrp.shape)
scatterPlot(TRXTsrp, TRy, 'Sparse Random Projection')

Wall time: 692 ms
(50000, 519)


from sklearn.manifold import Isomap
n_neighbors = 5
n_components = 10
n_jobs = 4

Eimap = Isomap(n_neighbors = n_neighbors,
              n_components = n_components,
              n_jobs=n_jobs)

%time Eimap.fit(TRXO)
%time TRXTimap = Eimap.transform(TRX)
TRXTimap = pd.DataFrame(data=TRXTimap, index=iTR)

VLXTimap = Eimap.transform(VLX)
VLXTimap = pd.DataFrame(data=VLXTimap, index=iVL)

print(TRXTimap.shape)
scatterPlot(TRXTimap, TRy, 'isomap')

Wall time: 620 ms
Wall time: 8.11 s
(50000, 10)


from sklearn.manifold import MDS

n_components = 3
n_init = 2
max_iter = 1200
metric = True
n_jobs = 4
random_state = 2018

Emds = MDS(n_components=n_components,
          n_init=n_init,
          max_iter=max_iter,
          metric=metric,
          n_jobs=n_jobs,
          random_state=random_state)

%time TRXTmds = Emds.fit_transform(TRXO)
TRXTmds = pd.DataFrame(data=TRXTmds, index=iTRO)

print(TRXTmds.shape)
scatterPlot(TRXTmds, TRy0, 'MDS')

Wall time: 1min 11s
(1000, 3)


TRXTmds


from sklearn.manifold import LocallyLinearEmbedding

n_neighbors = 10
n_components = 3
ethod = 'modified'
n_jobs = 4
random_state = 2018

Elle = LocallyLinearEmbedding(n_neighbors=n_neighbors,
                             n_components = n_components,
                             method='modified',
                             random_state=random_state,
                             n_jobs=n_jobs)

%time Elle.fit(TRX.loc[0:5000, :])
TRXTlle = Elle.transform(TRX)
TRXTlle = pd.DataFrame(data=TRXTlle, index=iTR)

VLXTlle = Elle.transform(VLX)
VLXTlle = pd.DataFrame(data=VLXTlle, index=iVL)

print(TRXTlle.shape)
scatterPlot(TRXTlle, TRy, 'LLE')

Wall time: 9.52 s
(50000, 3)


Elle.embedding_

array([[-0.00111031, -0.00486779, -0.00313505],
       [-0.03931153,  0.0089401 ,  0.0206251 ],
       [ 0.00713626,  0.02153579, -0.00420588],
       ...,
       [ 0.01072742, -0.02356577,  0.01404733],
       [ 0.00227599, -0.00813298,  0.00379371],
       [ 0.00834223,  0.01594336, -0.0012472 ]])


fig, axs = plt.subplots(nrows=2, ncols=2, figsize=(12, 12))
d0 = sns.scatterplot(x=TRXTlle.loc[:, 0], y=TRXTlle.loc[:, 1], hue=TRy==0, alpha=0.25, ax=axs[0,0]);
d0.set(xlabel='LLE1', ylabel='LLE2', title='0')
d1 = sns.scatterplot(x=TRXTlle.loc[:, 0], y=TRXTlle.loc[:, 1], hue=TRy==1, alpha=0.25, ax=axs[0,1]);
d1.set(xlabel='LLE1', ylabel='LLE2', title='1')
d5 = sns.scatterplot(x=TRXTlle.loc[:, 0], y=TRXTlle.loc[:, 1], hue=TRy==5, alpha=0.25, ax=axs[1,0]);
d5.set(xlabel='LLE1', ylabel='LLE2', title='5')
dd = sns.scatterplot(x=TRXTlle.loc[:, 0], y=TRXTlle.loc[:, 1], hue=TRy, alpha=0.25, ax=axs[1,1]);
dd.set(xlabel='LLE1', ylabel='LLE2', title='0~9');


TRXTO = TRXTlle.groupby(TRy).apply(lambda x:x.sample(100, random_state=2018))
iTRO = TRXTO.index.to_frame()[1]
TRXTO.set_index(iTRO, inplace=True)
TRyO = TRy.iloc[iTRO]

fig = px.scatter_3d(TRXTO, x=0, y=1, z=2,
                   color=TRyO==0,
                   symbol=TRyO==0,
                   size=1+(TRyO==0).astype(int),
                   size_max=9,
                   opacity=0.5)
fig.show()


fig = px.scatter_3d(TRXTlle, x=0, y=1, z=2, color=TRy, size_max=10)
fig.show()


from sklearn.manifold import TSNE

n_components = 3
learning_rate = 300
perplexity = 30
early_exaggeration = 12
init = 'random'
random_state = 2018

Etsne = TSNE(n_components=n_components,
            learning_rate=learning_rate,
            perplexity=perplexity,
            early_exaggeration=early_exaggeration,
            init=init,
            random_state=random_state)

%time TRXTtsne = Etsne.fit_transform(TRXTpca.iloc[:5000, :9])
TRXTtsne = pd.DataFrame(data=TRXTtsne, index=iTR[:5000])

print(TRXTtsne.shape)
scatterPlot(TRXTtsne, TRy, 't-SNE')

Wall time: 1min 20s
(5000, 3)


TMPy = TRy.loc[TRXTtsne.index]
fig = px.scatter_3d(TRXTtsne, x=0, y=1, z=2, color=TMPy, size=TMPy+1, size_max=8)
fig.show()


fig = px.scatter_3d(TRXTtsne, x=0, y=1, z=2, color=(TMPy==0), size=(TMPy==0)+1, size_max=8)
fig.show()


from sklearn.decomposition import MiniBatchDictionaryLearning

n_components = 50
alpha = 1
batch_size = 200
n_iter = 1000
random_state = 2018

Edl = MiniBatchDictionaryLearning(n_components = n_components,
                                 alpha=alpha,
                                 batch_size=batch_size,
                                 n_iter=n_iter,
                                 random_state=random_state)

%time TRXTdl =Edl.fit_transform(TRX)
TRXTdl = pd.DataFrame(data=TRXTdl, index=iTR)

VLXTdl = Edl.transform(VLX)
VLXTdl = pd.DataFrame(data=VLXTdl, index=iVL)

print(TRXTdl.shape)
scatterPlot(TRXTdl, TRy, 'Mini-batch Dictionary Learning')

C:\Users\wngus\anaconda3\envs\tf\lib\site-packages\sklearn\utils\validation.py:63: RuntimeWarning:

 Orthogonal matching pursuit ended prematurely due to linear
dependence in the dictionary. The requested precision might not have been met.

Wall time: 5min 7s
(50000, 50)


from sklearn.decomposition import FastICA

n_components = 25
algorithm = 'parallel'
whiten = True
max_iter = 100
random_state = 2018

Eica = FastICA(n_components = n_components,
              algorithm = algorithm,
              whiten = whiten,
              max_iter = max_iter,
              random_state = random_state)

%time Eica.fit(TRX)
TRXTica = Eica.fit_transform(TRX)
TRXTica = pd.DataFrame(data=TRXTica, index=iTR)

VLXTica = Eica.transform(VLX)
VLXTica = pd.DataFrame(data=VLXTica, index=iVL)

scatterPlot(TRXTica, TRy, 'Independent Component Analysis')

Wall time: 15.9 s


fig = px.scatter_3d(TRXTica, x=0, y=1, z=2, color=(TRy==2), size=(TRy==2)+1, size_max=8)
fig.show()


print("--- %s seconds ---" % (time.time() - start_time))

--- 1034.4201645851135 seconds ---

	0	1	2	9	19	49	99	199	299	399	499	599	699
Eigenvalue	5.1083	3.7010	3.2587	1.2403	0.6034	0.1683	0.0528	0.0148	0.0071	0.0031	0.0008	0.0001	0.0
Prop	0.0974	0.0706	0.0622	0.0237	0.0115	0.0032	0.0010	0.0003	0.0001	0.0001	0.0000	0.0000	0.0
Cumulative	0.0974	0.1680	0.2302	0.4888	0.6440	0.8249	0.9147	0.9665	0.9862	0.9958	0.9993	1.0000	1.0

	x351	x407
0	0.273438	0.937500
1	0.000000	0.000000
2	0.000000	0.941406
3	0.800781	0.980469
4	0.031250	0.988281

	mean_fit_time	std_fit_time	mean_score_time	std_score_time	param_n_components	params	split0_test_score	split1_test_score	split2_test_score	split3_test_score	split4_test_score	mean_test_score	std_test_score	rank_test_score
0	4.002801	1.024797	0.395897	0.136476	100	{'n_components': 100}	-0.005773	-0.005762	-0.005747	-0.005782	-0.005752	-0.005763	1.288494e-05	5
1	4.708631	0.249394	0.301810	0.012087	200	{'n_components': 200}	-0.002301	-0.002288	-0.002290	-0.002300	-0.002289	-0.002293	5.644258e-06	4
2	6.910282	0.320145	0.350451	0.016190	300	{'n_components': 300}	-0.000956	-0.000953	-0.000955	-0.000958	-0.000953	-0.000955	2.186464e-06	3
3	9.537575	0.248477	0.423061	0.034302	400	{'n_components': 400}	-0.000292	-0.000290	-0.000291	-0.000293	-0.000290	-0.000291	1.104186e-06	2
4	13.013720	0.612896	0.442413	0.045059	500	{'n_components': 500}	-0.000054	-0.000053	-0.000053	-0.000054	-0.000054	-0.000054	5.174543e-07	1

	mean_fit_time	std_fit_time	mean_score_time	std_score_time	param_gamma	param_kernel	params	split0_test_score	split1_test_score	split2_test_score	mean_test_score	std_test_score	rank_test_score
0	0.115216	0.004526	0.045042	0.004480	0.01	rbf	{'gamma': 0.01, 'kernel': 'rbf'}	-0.062472	-0.056276	-0.055742	-0.058163	0.003054	8
1	0.099569	0.002432	0.035241	0.000939	0.01	linear	{'gamma': 0.01, 'kernel': 'linear'}	-0.055510	-0.055971	-0.053476	-0.054986	0.001084	1
2	0.115672	0.005090	0.042455	0.001034	0.02	rbf	{'gamma': 0.02, 'kernel': 'rbf'}	-0.059291	-0.053214	-0.052648	-0.055051	0.003007	6
3	0.095320	0.000929	0.033658	0.001276	0.02	linear	{'gamma': 0.02, 'kernel': 'linear'}	-0.055510	-0.055971	-0.053476	-0.054986	0.001084	1
4	0.113283	0.001937	0.046228	0.007153	0.03	rbf	{'gamma': 0.03, 'kernel': 'rbf'}	-0.061458	-0.055114	-0.054479	-0.057017	0.003151	7
5	0.106218	0.006000	0.035902	0.002726	0.03	linear	{'gamma': 0.03, 'kernel': 'linear'}	-0.055510	-0.055971	-0.053476	-0.054986	0.001084	1
6	0.113543	0.002017	0.042211	0.001231	0.04	rbf	{'gamma': 0.04, 'kernel': 'rbf'}	-0.064699	-0.058224	-0.057510	-0.060144	0.003234	9
7	0.098913	0.002717	0.034059	0.000881	0.04	linear	{'gamma': 0.04, 'kernel': 'linear'}	-0.055510	-0.055971	-0.053476	-0.054986	0.001084	1
8	0.114649	0.008092	0.043688	0.003793	0.05	rbf	{'gamma': 0.05, 'kernel': 'rbf'}	-0.067688	-0.061154	-0.060380	-0.063074	0.003278	10
9	0.102030	0.008127	0.040293	0.003676	0.05	linear	{'gamma': 0.05, 'kernel': 'linear'}	-0.055510	-0.055971	-0.053476	-0.054986	0.001084	1

	0	1	2
1
1907	3.742804	-5.522545	5.454732
5809	2.836946	-6.351985	-0.773530
32846	-0.034010	-4.593104	8.550494
8838	-0.778592	-3.789294	8.203202
13044	-1.427745	-1.683926	8.627664
...	...	...	...
5459	5.406290	0.307090	-3.424906
34983	4.955261	1.458435	-2.962013
16622	0.360030	2.772790	-4.439113
32504	-1.408353	6.400787	3.725603
27565	3.446744	-0.049185	-3.483470

자료읽기¶

개별 이미지 시각화¶

레이블 조정하기¶

층화추출한 자료¶

PCA¶

모델 적합/성분계산¶

모델 설명력¶

차원축소 Tuning¶

Incremental PCA¶

Sparse PCA¶

Kernel PCA¶

tuencatedSVD¶

GRP¶

Sparse Random Projection¶

Isomap¶

MDS¶

LLE¶

t-SNE¶

MiniBatch Dictionary Learning¶

ICA¶

	x01	x02	x03	x04	x05	x06	x07	x08	x09	x10	...	x775	x776	x777	x778	x779	x780	x781	x782	x783	x784
count	50000.0	50000.0	50000.0	50000.0	50000.0	50000.0	50000.0	50000.0	50000.0	50000.0	...	50000.000000	50000.000000	50000.000000	50000.000000	50000.000000	50000.000000	50000.0	50000.0	50000.0	50000.0
mean	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000739	0.000354	0.000204	0.000090	0.000071	0.000009	0.0	0.0	0.0	0.0
std	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.022784	0.015424	0.012080	0.007218	0.007181	0.001483	0.0	0.0	0.0	0.0
min	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.0	0.0	0.0	0.0
25%	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.0	0.0	0.0	0.0
50%	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.0	0.0	0.0	0.0
75%	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.0	0.0	0.0	0.0
max	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.992188	0.992188	0.988281	0.988281	0.992188	0.242188	0.0	0.0	0.0	0.0

	0	1	2	3	4	5	6	7	8	9	...	774	775	776	777	778	779	780	781	782	783
0	1.0	-0.0	0.0	-0.0	0.0	0.0	-0.0	0.0	-0.0	0.0	...	-0.00000	-0.00000	0.00000	0.00000	-0.00000	0.00000	0.00000	0.00000	0.00000	-0.00000
1	-0.0	1.0	-0.0	-0.0	-0.0	-0.0	0.0	0.0	0.0	0.0	...	0.00000	0.00000	0.00000	0.00000	-0.00000	-0.00000	-0.00000	-0.00000	-0.00000	0.00000
2	0.0	-0.0	1.0	0.0	-0.0	-0.0	-0.0	0.0	-0.0	0.0	...	0.00000	-0.00000	0.00000	-0.00000	-0.00000	-0.00000	-0.00000	-0.00000	-0.00000	0.00000
3	-0.0	-0.0	0.0	1.0	-0.0	0.0	-0.0	0.0	0.0	0.0	...	-0.00000	-0.00000	-0.00000	0.00000	0.00000	-0.00000	-0.00000	-0.00000	0.00000	0.00000
4	0.0	-0.0	-0.0	-0.0	1.0	0.0	-0.0	0.0	0.0	0.0	...	0.00000	-0.00000	-0.00000	0.00000	-0.00000	0.00000	0.00000	0.00000	0.00000	-0.00000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
779	0.0	-0.0	-0.0	-0.0	0.0	0.0	0.0	0.0	0.0	-0.0	...	0.00004	0.00003	0.00003	0.00003	0.00003	1.00000	-0.00002	0.00004	0.00002	-0.00003
780	0.0	-0.0	-0.0	-0.0	0.0	-0.0	0.0	0.0	0.0	-0.0	...	0.00001	0.00001	0.00001	0.00001	0.00001	-0.00002	1.00000	0.00001	0.00001	-0.00001
781	0.0	-0.0	-0.0	-0.0	0.0	-0.0	0.0	0.0	0.0	-0.0	...	-0.00002	-0.00002	-0.00002	-0.00002	-0.00002	0.00004	0.00001	1.00000	-0.00001	0.00002
782	0.0	-0.0	-0.0	0.0	0.0	-0.0	-0.0	-0.0	-0.0	-0.0	...	-0.00001	-0.00001	-0.00001	-0.00001	-0.00001	0.00002	0.00001	-0.00001	1.00000	0.00001
783	-0.0	0.0	0.0	0.0	-0.0	0.0	0.0	0.0	0.0	0.0	...	0.00002	0.00002	0.00002	0.00002	0.00002	-0.00003	-0.00001	0.00002	0.00001	1.00000

	0	1	2	3	4	5	6	7	8	9	...	774	775	776	777	778	779	780	781	782	783
0	5.10829	-0.00000	0.00000	-0.00000	-0.00000	0.0	-0.0	0.0	-0.0	0.0	...	-0.0	0.0	-0.0	-0.0	-0.0	0.0	0.0	0.0	0.0	-0.0
1	-0.00000	3.70098	-0.00000	-0.00000	-0.00000	-0.0	0.0	-0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	-0.0	-0.0	-0.0	-0.0	-0.0	0.0
2	0.00000	-0.00000	3.25868	0.00000	-0.00000	-0.0	-0.0	0.0	0.0	0.0	...	0.0	-0.0	0.0	-0.0	-0.0	-0.0	-0.0	-0.0	-0.0	0.0
3	-0.00000	-0.00000	0.00000	2.82008	-0.00000	0.0	-0.0	0.0	0.0	0.0	...	-0.0	-0.0	0.0	-0.0	0.0	-0.0	-0.0	-0.0	0.0	0.0
4	-0.00000	-0.00000	-0.00000	-0.00000	2.54673	0.0	-0.0	0.0	0.0	0.0	...	-0.0	-0.0	0.0	-0.0	0.0	0.0	0.0	0.0	0.0	-0.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
779	0.00000	-0.00000	-0.00000	-0.00000	0.00000	0.0	0.0	0.0	0.0	-0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	-0.0	0.0	0.0	-0.0
780	0.00000	-0.00000	-0.00000	-0.00000	0.00000	-0.0	0.0	0.0	0.0	-0.0	...	0.0	0.0	0.0	0.0	0.0	-0.0	0.0	0.0	0.0	-0.0
781	0.00000	-0.00000	-0.00000	-0.00000	0.00000	-0.0	0.0	0.0	0.0	-0.0	...	-0.0	-0.0	-0.0	-0.0	-0.0	0.0	0.0	0.0	-0.0	0.0
782	0.00000	-0.00000	-0.00000	0.00000	0.00000	-0.0	-0.0	-0.0	-0.0	-0.0	...	-0.0	-0.0	-0.0	-0.0	-0.0	0.0	0.0	-0.0	0.0	0.0
783	-0.00000	0.00000	0.00000	0.00000	-0.00000	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	-0.0	-0.0	0.0	0.0	0.0

	n
TRy	50000
VLy	10000
TSy	10000